IPUToolkit.jl
IPUToolkit.jl copied to clipboard
IPU programming in Julia
This pull request changes the compat entry for the `LLVM` package from `6` to `6, 7`. This keeps the compat entries for earlier versions. Note: I have not tested your...
In some cases, e.g. if no IPU devices are available for some reasons (including misconfiguration of the current system), `Poplar.get_ipu_device()` may return a device, which is actually an IPU model,...
Clang.jl is currently not compatible with Julia v1.11, so we can't test this at the moment. Fixing #18 would let us reduce the number of dependencies, including Clang.jl itself (it's...
Continuation of #37. Proof of concept for #18. Still lots to do: * [x] the resulting library is only 21 kB, it's definitely wrong, it should be over ~40 MB,...
This pull request changes the compat entry for the `CxxWrap` package from `0.14` to `0.14, 0.15`. This keeps the compat entries for earlier versions. Note: I have not tested your...
CC: @maleadt
Currently we only provide the macro [`@codelet`](https://juliaipu.github.io/IPUToolkit.jl/stable/compiler/#IPUToolkit.IPUCompiler.@codelet) to build the `.gp` codelet _and_ add it to a graph, but it was suggested that it can be useful to just compile...
``` struct MyVertex compute(); } ``` We could codegen that `__runCodlet` function. `@llvm.colossus.get.vertex.base()` is the intrinsic to get the vector base. The big unknown is how to do the mapping...